Audio Hiss Problem--I tried audacity

[bbbbbbb45]

I'm sure I'm not the first to ask this question, but I haven't been able to find a solution online.

I get a background hiss problem when I film. I tried using Audacity to remove it, but the more of the hiss I remove, the "flatter" or more muffled the audio becomes. I have also heard, anecdotally, that goldwave is no better at this, but I haven't tried it.

1) Is there any better software solution to removing hiss without sacrificing much audio quality?

2) I'm using $25 lavalier mikes--can someone recommend a microphone solution that doesn't cost thousands of dollars?
Thanks
K

[jagabo]

The biggest issue with microphones is to get the mic as close to the speaker (or whatever is making the sounds you want to hear) as possible.

Did you try the Noise Reduction filter in Audacity? Train it with as "silent" section then remove noise. It usually works best if you run it a few times with moderate reductions (retraining each time) rather than one time with a big reduction.

[bbbbbbb45]

Hi. The microphone is lavalier, meaning it is clipped onto the person using it. I have tried Audacity as I mentioned but the loss in audio quality was not very good.

[jagabo]

I realize a lavalier is clipped to the person. I was pointing out that it's better to have it clipped to their collar than lower on their shirt. Regarding Audacity, it has lots of filters. I was pointing out that you should use the Noise Reduction filter and how to use it, not say, a graphic equalizer filter. I can't read your mind and know what you tried.

[bbbbbbb45]

I appreciate your response, but have tried what you mentioned (specifically the NR filter in Audacity) and still experience the loss of sound quality.
--K

[jagabo]

You always get a loss of sound quality when you reduce noise. If you post a sample of your source some people here will probably give you some more suggestions.

[Cornucopia]

1. GET A BETTER MIKE! $25? that's bottom of the barrel. You don't have to pay $1k+ to get good sound, but you DO have to pay more than $25. Try something like an AT, Sennheiser, Sony, or Shure mike (I'm assuming a WIRED lav here, right?). Price ranges could be between $150-$500 or more. Read reviews on audio boards... (and tell us more specifically your intended uses, environments & budget)

2. Make your mike AS CLOSE AS POSSIBLE to your sound source, and AS FAR AWAY AS POSSIBLE from your interfering sounds (this can include the camera/you, lights, and the environment).

3. Get a mike that is sensitive in mainly ONE direction (aka "Unidirectional" or one of the "Cardioid" family) as opposed to sensitive to ALL directions (aka "Omnidirectional"). Then POINT the mike in the right direction.

4. Make sure you have correct levels (aka "gain structure"), so that you are using the most of the available dynamic range inherent in the mike. IOW, if you have a mike with a "weak" or "strong" level switch, and you set it to "weak" and then have to crank up the gain sensitivity in order to get a loud enough sound under usual circumstances, you're doing yourself a disservice and ADDING noise.

5. That last part includes using the right interfaces (Pro vs. Consumer, XLR vs. Phono, Balanced vs. Unbalanced, etc) and using a proper converter if your interfaces don't natively match.

6. Do level tests to see what your mike CAN put out, and what your camera can TAKE so that you get the best match consistently.

7. Monitor the signal with headphones so you know WYHearIWYG.

All those dealt with options BEFORE it was recorded. At that point, it should already sound either OK, Good, Great, or Fantastic. If it doesn't sound like one of those, you're NOT following one or more of those rules...

But if it only sounds OK or Good, and you want it better:

8. Use the proper filtering in Audio Editor apps...
A. Bandpass filter set to ~90Hz up to 5-8kHz gets rid of stuff that isn't in the human vocal range (that's for Talking, not Singing - extend if Singing).
B. Dynamic range compression + Peak limiting + Low level expansion/Noise gate combined and using medium fast attack and slower release will work to bring out the power/presence of the voice compared to other sounds.
C. If there's still noise, you could do Broadband Noise Reduction where you "learn" the frequency spectrum character of the noise (by sampling a section where there is JUST noise) and then digitally subtract it from the overall signal. BEWARE! This is a strong tool that can easily be overdone, resulting in just the kind of "loss of clarity" that you opined.

There are a few other things to try, but that should get you going.

One last thing to remember, keep a keen ear out for what is actually going on during the recording, because if it actually ends up where the NOISE/BACKGROUND is louder than the signal you wanted, NO AMOUNT of processing will EVER get you something satisfactory.

Scott

[bbbbbbb45]

Good idea. I have uploaded a sample mp3. When I try to eliminate the noise, the voice gets "flattened"
Thanks
K

[jagabo]

You can't do serious noise reduction with an 128 kbps MP3 file, it's too screwed up. You should upload an uncompressed WAV file or flac of the original source. Unless MP3 is all you have.

[bbbbbbb45]

Here's the wav version
Thanks

[jagabo]

I'm not especially good at this but here's your sample with -10dB noise reduction (two passes of Noise Removal, -5 dB each) and a little high end boost to sharpen it up. You can go with more noise removal but the hiss starts pumping noticeably.

[mike20021969]

Here's my attempt.

[poisondeathray]

Your money is probably better spent getting a "clean" recording in the first place, than trying to "fix" it after the fact.

I find iZotope has one of the better NR algorithms (compared to other commonly used audio software like soundforge, audition/soundbooth)

[davexnet]

"test2.wav" doesn't sound too bad to me as it is. I must admit, I don't have the best listening scenario.
Audio interference from the fans in my PC (which is raised up onto the table close to the monitor), or the frequency
response of the onboard audio/speakers. Of course, it could be my hearing...

[bbbbbbb45]

Thanks for all your feedback!!!!!
1) Mike, I appreciate your efforts, but your sample sounded as muffled as the ones I attempt
2) P,
a) you gave two samples, the 18 one was the best--what settings did you use?
b) Also, did you use Izotop's rx2, anr-b, or the cheaper music & speech cleaner, and is there a difference between the three for this purpose?

3) Jagabo's sample wasn't perfect but was a substantial improvement. Exactly what kind of "high end boost" did you give it (what is a high end boost in audacity?)

4) Cornu (thanks for your tips!) and others, do you have any specific recommendations for lavalier mikes that DON'T have audio hiss in the $100-$300 range that work with camcorders?

Thanks for all your time!!!!
K

[jagabo]

Exactly what kind of "high end boost" did you give it (what is a high end boost in audacity?)

I used the graphics eq to add a 3dB boost between about 5 and 10 KHz. That restored crispness to the sybilants, getting the clarity very close to the original. It also increased the noise in that region a bit but I thought that was a reasonable tradeoff.

[poisondeathray]

2) Poison,
a) you gave two samples, the 18 one was the best--what settings did you use?

Default "slow" settings profile. Nothing special.

The only adjustment I made was the amount of dB removal. However, there are many other advanced settings that you can tweak in the NR removal module if you are so inclined

b) Also, did you use Izotop's rx2, anr-b, or the cheaper music & speech cleaner, and is there a difference between the three for this purpose?

RX2 . I'm not familiar with the other products